Computers in Human Behavior 26 (2010) 226–234
Contents lists available at ScienceDirect
Computers in Human Behavior
journal homepage: www.elsevier.com/locate/comphumbeh
Face off: Implications of visual cues on initiating friendship on Facebook
Shaojung Sharon Wang *, Shin-Il Moon, Kyounghee Hazel Kwon, Carolyn A. Evans, Michael A. Stefanone
Department of Communication, University at Buffalo, The State University of New York, 359 Baldy Hall, Buffalo, NY 14260-1020, USA article info abstract
Article history: This research investigates how moderating factors and theoretically relevant contextual variables affect Available online 31 October 2009 impression formation and the willingness to initiate virtual friendship. An experiment examined both main and interaction effects for visual cues, profile owner’s gender, and evaluator’s gender; a 2 (stimulus Keywords: gender: male and female) Â 3 (visual conditions: attractive, unattractive, and no-photo) Â 2 (evaluator’s CMC gender: male and female) between subjects model analysis of variance (ANOVA) was employed. A three- Facebook way interaction between gender and appearance was revealed. The results indicated that both male and Visucal cues female subjects were more willing to initiate friendships with opposite-sex profile owners with attractive Impression formation photos. Subjects also displayed comparatively higher willingness to make friends with profile owners Physical attractiveness Hyperpersonal who did not include visual cues than with those who revealed an unattractive photo. The hyperpersonal model was supported and extended to address gender attributes. Ó 2009 Elsevier Ltd. All rights reserved.
1. Introduction mental requirement when initiating friendships. Gathering such information offline typically involves only a few important cues The Internet has become a principal venue for social interaction. for perceivers to form impressions of their target communication There is a growing body of literature suggesting that the Internet partners and physical appearance is the most accessible of these and its communication tools facilitate the maintenance of existing cues (Jones, 1990; Kenny, 1991). Visual cues are extremely impor- relationships (Katz & Rice, 2002), formation of romantic connec- tant as they function to form initial impressions between zero-his- tions (McKenna, Green, & Gleason, 2002), construction of virtual tory dyads quickly and accurately (Ambady & Rosenthal, 1992) community (Rheingold, 1996), and most notably, the emergence that can impact the willingness to build future relationships. While of online friendships (Parks & Floyd, 1995). In light of the wide- there is ample evidence supporting this proposition offline, little spread use of social network sites (SNSs) in the last few years, there research has examined the role of these visual cues in interper- has been interest in the potential of the Internet to create new rela- sonal relationship formation online. tionships; Facebook.com is an example of a SNS where people Computer-mediated communication (CMC) places emphasis on communicate and foster friendship with each other in cyberspace. reduced communication cues, synchronous or asynchronous com- In August 2009, the number of American visitors to Facebook munication (Walther, 1996), and anonymity (McKenna & Bargh, reached more than 90 million and it was ranked 5th on the Top 2000). The social identification model of deindividuation effects 50 US Web Properties list (comScore, 2009). As Facebook has fo- (SIDE) (Lea & Spears, 1991, 1995) posits that the anonymity and cused its attention on expansion and increased adoption world- lack of identifiability inherent in most CMC serves to accentuate wide, it is now the fourth largest site in the world; with 340 the effect of the salient social identity and strengthen the domi- million unique global visitors as of June, 2009 (Schonfeld, 2009). nant normative response associated with it. The SIDE model fo- Hitwise report also revealed that Facebook and other popular SNSs cuses on the cognitive processes by which CMC communicators have overtaken adult entertainment as the most visited website use social identity variables and social category information to category on the Internet for 18 to 24 year-olds (Tancer, 2008). make inferences and over- environment should result in deindivid- Facebook allows users to create profiles that often contain per- uation which leads to increasingly stereotyped and exaggerated sonal information including photos and video images. It is a vehicle representations of others (Hancock & Dunham, 2001). Given the for posting and sharing visual biographies to maintain friendships deindividuating features of CMC and exaggerated stereotypic rep- with acquaintances and for exploring relationships with formerly resentations in asynchronous settings, CMC can even surpass the unknown others. Having access to personal information is a funda- level of affection and emotion of parallel face-to-face (FtF) interac- tion (Walther, 1996). Early CMC did not support facial expression or verbal intonation * Corresponding author. Tel.: +1 716 645 2141. as a means for establishing social presence (Short, Williams, & E-mail addresses: [email protected], [email protected] (S.S. Christie, 1976). This visual anonymity allowed CMC users to Wang).
0747-5632/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.chb.2009.10.001 S.S. Wang et al. / Computers in Human Behavior 26 (2010) 226–234 227 minimize undesirable physical or behavioral cues and select more 82% of them use SNS to keep in touch with friends they rarely favorable information to disclose; selective display of these cues see in person (but already know). This study also found, however, enables the deliberate presentation of self in early stages of rela- that 49% of SNS users (27% of online teens) use these sites to make tionship development (Goffman, 1963). This is not the case today, new friends. More specifically, the Pew study showed that nearly however, as both visual and vocal cues are no longer rare in con- half (47%) of social network-using older teen boys (ages 15–17) temporary CMC settings. For example, research on gender stereo- had friends in their SNS who they had never met, as did 28% of old- typing in a CMC environment reveals that gendered cartoon er teen girls. Further, Stefanone, Lackaff, and Rosen (2008) found characters (visual cues) yield stronger effects on gender inferences on average about 15% of Facebook networks were comprised of and conformity behavior than do verbal cues (Lee, 2007). More re- people who had never actually met in person. cent research on online dating (Hancock & Toma, 2009) examined The new relationship formation process for the Facebook gener- photos posted on online dating profiles and the extent to which the ation has raised questions about impression management and self- photo resembles its owner’s typical appearance. This study high- presentation in this less anonymous online environment. Zhao, lights the importance of online self-presentation of one’s physical Grasmuck, and Martin (2008) found that the visual possibilities attractiveness through visual elements like photographs. of Facebook enable users to showcase themselves implicitly On SNSs like Facebook photographs are selectively displayed to through their friend list, photos, and wall postings and thus require create positive self-presentations for potential relationships. The audiences’ equal attention to their social milieu. In this way, they selective presentation of self and anticipated future interaction argued users create hoped-for ‘‘digital selves” to enhance their characteristic of Facebook users raises many questions. For exam- overall self-images and chances to connect in the offline world. ple, how do visual cues affect relationship formation online? Are Likewise, the number of friends Facebook users collect (Tong, visual cues that are important in offline relationships also impor- Van Der Heide, Langwell, & Walther, 2008) and the physical attrac- tant in the initiation of virtual friendships? Whereas people may tiveness and wall postings of their friends shown on their profiles not substitute their traditional social networks for online relation- (Walther, Van Der Heide, Kim, Westerman, & Tong, 2008) provide ships maintained exclusively via Facebook, this study does not meaningful cues to shape observers’ impressions of profile owners. examine what constitutes online friendship. Nor does it focus on This line of research, derived from Walther’s (1992) social informa- the extent to which online interactions result in offline relation- tion processing (SIP) theory, employs a Brunswikian Lens Model ships. Rather, the goal of this study is to achieve a better under- (see Walther et al., 2008, for review) on judgments about uncer- standing of the theoretical mechanisms at play when people tainty to explore how the attractiveness of friends’ photos and text make decisions about whom to become friends with via SNSs. In shown on Facebook profiles can impact the profile owners’ per- addition we propose to extend the hyperpersonal model by sys- ceived attractiveness. These cues influence judgments about pro- tematically analyzing how moderating factors and theoretically file owners even though theprofile owners themselves did not relevant contextual variables affect impression formation and the provide that information (Walther et al., 2008). willingness to initiate friendship online. Most of the research to date on Facebook has focused on con- necting and reconnecting people, as well as addressing privacy and self-disclosure concerns. This new approach, however, illus- 2. Research on Facebook trates the process of electronic impression formation by newer communication technologies and highlights the warrant of SIP the- Numerous empirical studies have been conducted to examine ory. That is, people use whatever information is available to make patterns of college students’ use of Facebook. These focused on a judgments about their online partners and impression formation variety of academic interests including characteristics of profile about others encountered online is the same process used for oth- elements (Lampe, Ellison, & Steinfield, 2007), self-presentation ers encountered offline. (Stutzman, 2006), surveillance and privacy concerns (Gross & Although there is no direct relationship between the number of Acquisti, 2005; Joinson, 2008; Peluchette & Karl, 2008), social cap- friends SNS users have and their perceived attractiveness by oth- ital (Ellison, Steinfield, & Lampe, 2007) and social grooming (Tufek- ers, the number of friends may function as a proxy for physical ci, 2008), relationship marketing strategies for the Facebook attractiveness when photographic cues are absent (Tong et al., generation (Meadows-Klue, 2008), and students’ perceptions of 2008). By the same token, the extent text left by one’s friends con- teacher self-disclosure via Facebook (Mazer, Murphy, & Simonds, tributes to the perception of physical attractiveness when photo- 2007). While some studies indicate negative outcomes of Facebook graphs are not available remains unknown. Indeed even though use (e.g., stalking and identity theft), others demonstrate how both visual and textual cues are now accommodated in SNSs, the Facebook can generate a variety of positive social outcomes such implications of visual cues on making friends online have not been as enhanced social capital and collaboration. fully explored. Further, factors involved with initiating friendship One general conclusion of these studies was that Facebook in CMC environments have not been extensively studied. A brief plays a significant role in supporting pre-existing social relations. review of past CMC as well as interpersonal communication liter- For example, when Ellison et al. (2007) examined whether or not ature can further illuminate the scope of these issues and their Facebook is used to meet new people, the authors found that col- investigation. lege students use Facebook primarily to maintain relationships with shared social groups (e.g., high-school classmates, old friends, and current classmates) or a shared location (e.g., dormitory). This 3. Theoretical framework differed from early CMC research on virtual communities which posited that online community members would be more likely to Based on the idea that communication media’s social effects dif- meet others outside their pre-existing social group or location fer in the degree to which they allow people to experience others (Wellman, Salaff, Dimitrova, Yuan, & Gay, 2006). Facebook thus as being psychologically present by providing a sense of intimacy generally supports geographically-bound communities exhibiting and immediacy (Short et al., 1976), the core of early CMC research a tendency to flow from offline to online interactions. (Culnan & Markus, 1987; Daft and Lengel, 1986; Sproull & Kiesler, This tendency was also supported by a recent Pew Internet and 1986) was founded on examining variability in the capacity of CMC American Life Project Report (Lenhart & Madden, 2007) indicating to transmit social context cues. The cues-filtered-out (CFO) ap- 91% of US teens use SNSs to contact friends they see often while proach (Culnan & Markus, 1987) combines social presence theory 228 S.S. Wang et al. / Computers in Human Behavior 26 (2010) 226–234
(Short et al., 1976) and lack of physical and salient social cues simulated Facebook profiles between six conditions: photo (no- framework to posit that CMC impedes relationship development photo, attractive photo, unattractive photo) Â sex (female/male due to the reduction in non-verbal cues. This perspective high- profile owners). lights the apparent need for physical/visual appearance in effective communication as the amount of information between and the le- 3.2.1. Display of visual cues vel of affection toward CMC partners are positively related (Storck Past studies of interpersonal communication and friendship for- & Sproull, 1995). mation processes (Duck, 1982; Udry, 1966) operationalized the Visual cues which are more dominant than verbal cues are ab- ‘‘acquaintance stage” (Duck, 1982, p. 135) in which people first sent in traditionally text-based CMC. Contemporary Internet-based encounter, then react to and make impressions about their for- communication tools like SNSs, however, were designed specifi- merly unacquainted counterparts. During this stage, people make cally to accommodate visual cues including images and videos. judgments about each other prior to dyadic interaction. Here they These mediated environments facilitate synchronous display of primarily rely on visual cues like attractiveness, race, height, and both visual and textual cues; this contrasts sharply with early hair color to evaluate the viability of new relationships. Visual cues text-based CMC as visual cues now serve as prominent elements are critically important during the acquaintance stage; only after during impression formation. this stage do people consider deeper and more cognitive informa- tion such as personality similarity (Duck, 1982). 3.1. The hyperpersonal model While visual cues are important during early stage interaction, facial characteristics are not only the main source of most non-ver- Contrary to the CFO approach, evidence suggests there is a high bal information but also an essential source of information about degree of socioemotional content in CMC (Reingold, 1993) and emotion, age, race, and sex (Alley, 1988; Ekman, 1992). When peo- although there are fewer paralinguistic cues, people who are ‘‘sea- ple perceive and evaluate others, they tend to look at their faces. soned communicators” in CMC become ‘‘adept at using and inter- Prior to engaging in conversations, these facial characteristics preting textual signs and paralinguistic codes” (Lea & Spears, 1995, influence perceptions about the willingness to establish relation- p. 217). People tend to rely on a range of different cues like lan- ships with others as well. In this study we specifically manipulate guage, style, timing and speed of writing, punctuation, or emoti- Facebook profile owners’ portraits in which the face and its expres- cons when making inferences about potential relationships sion are predominant. Thus, we hypothesize. (Mantovani, 2001). This is congruent with the SIP perspective H1: There will be a main effect on display of profile photos (vi- (Walther, 1992) which posits that social exchange and subsequent sual cues) on Facebook. relationship development via CMC, given enough time, is equiva- lent to that of offline relationships. Hence, the presence of photo- graphic images of each other to promote affection and social 3.3. Physical attractiveness stereotype attraction is less effective for long-term CMC partners than it is for unacquainted CMC group interaction (Walther, Slovacek, & Tid- While SNSs allow users to present ideal selves, physical attrac- well, 2001). These findings resonate with the hyperpersonal model tiveness stereotypes (Dion, Berscheid, & Walster, 1972) operate to (Walther, 1996) which emphasizes people’s tendency to over-attri- influence the way people create preliminary impressions about bute available information to create an idealized image of others in others when visual cues are presented. People’s perception of their the absence of other cues during CMC. romantic partner’s physical attractiveness varies depending on the Walther (1996) argued that in the absence of non-verbal cues in quality of social interaction developed over time (Albada, Knapp, & lean media, message senders have a greater ability to strategically Theune, 2002). In the context of new relationships, however, phys- develop and selectively present themselves whereas message ical attractiveness functions as a predominant element to predict receivers idealize and build their impressions on any bit of infor- individuals’ attitude at the acquaintance stage (Berscheid & mation they receive about their communicative partners. In short, Walster, 1974). selective self-presentation combined with selective reception can Results from numerous studies on physical attractiveness indi- foster heightened senses of liking and intimacy. The hyperpersonal cate that those who are physically attractive reap greater social re- model, as Walther et al. (2001) explain, ‘‘depicts how senders se- wards ranging from more romantic dates to higher average lect, receivers magnify, channels promote, and feedback increases incomes (Feingold, 1992; Roszell, Kennedy, & Grabb, 1989). Attrac- enhanced and selective communication behaviors in CMC” (p. tive people are believed to possess a number of positive character- 110). istics (Riggio, 1986) and lead more fulfilling lives (Feingold, 1992). The hyperpersonal model clarifies how the presence of visual Dion et al. (1972) claim that, in people’s perception of others, cues like photographs affect interpersonal attraction for unac- ‘‘what is beautiful is good” (p. 285) and abundant research has doc- quainted CMC partners and further postulates that the greatest umented this stereotype (Adams, 1982; Alley & Hildebrandt, 1988; outcome of social interaction mainly occurs after a longer period Dion, 1981, 1986; Hatfield & Sprecher, 1986; Patzer, 1985)asa of time if text only is used alone by CMC users (Walther, 1996). prevailing phenomenon exhibited in everyday life. More precisely, it suggests that CMC communicators form exagger- The online profiles presented in this study contain limited per- ated perceptions about their partners while visual cues are absent sonal information that does not function to enhance perceived so- and makes predictions about what happens during ongoing inter- cial presence (Short et al., 1976). When the physical reality of the actions. It does not emphasize, however, the role of non-verbal other person in a mediated environment is absent, pictures can cues on once-off profile perceptions in initiating online relation- convey visual impressions that support the principle of cue substi- ships at the unacquainted stage. tutability to intimacy (Argyle & Dean, 1965) because subjects can adopt affective messages of the profile owners. The influence of 3.2. The current study physical attractiveness is likely as important in CMC as it is offline. Building on the large body of literature which suggests physical Because selective self-presentation is enhanced in text-based attractiveness is associated with interpersonal attraction, we posit CMC (Walther, 2007), the current study explores the importance H1a: Those who are exposed to attractive profile photos will be of displaying visual cues during the initial stage of online friend- more willing to initiate friendships than those in the unattractive ship formation. In this study, visual cues are examined through and no-photo conditions. S.S. Wang et al. / Computers in Human Behavior 26 (2010) 226–234 229
The hyperpersonal model suggests that reduced non-verbal Facebook users make friends with both sexes. Users display cues in CMC lead to an idealized perception by message receivers their own as well as their friends’ photos on their profiles. Research to inflate the impression and generalize positive cues on other un- has demonstrated that physical attractiveness of profile owners’ known personality aspects of message senders. We incorporate friends directly influence profile owners’ perceived attractiveness physical attractiveness stereotypes and hyperpersonal affinity to (Walther et al., 2008). The radiation effect may occur when having propose an attractive friend of the other sex on Facebook user’s profile. H1b: Those who are exposed to unattractive profile photos will Having a physically more desirable photo of a same-sex friend be less willing to initiate friendships than those exposed to the no- may, however, result in a ‘‘contrast effect” on the profile owner photo condition. during courtship. We predict that attractive people are more desir- able as opposite-sex virtual friends but are less preferred as same- sex virtual friends. Thus, the interaction between a profile owner’s 3.4. Physical attractiveness and gender gender and evaluator’s gender can be further qualified such that H3a: Evaluators will be more willing to initiate friendships with Literature on friendship formation suggests that opposite-sex opposite-sex profile owners with attractive photos than with and same-sex friendships represent a qualitatively different type attractive same-sex dyads. of interaction. Both men and women prefer opposite-sex friends H3b: Evaluators will be less willing to initiate friendships with than same-sex ones for obtaining ‘‘insider” information of oppo- opposite-sex profile owners with unattractive photos than with site-sex mates (Bell, 1981; Bleske & Buss, 2000; Hacker, 1981; Sap- unattractive same-sex dyads. adin, 1988). These preferences for opposite-sex friendships have Finally, the hyperpersonal model asserts that it is more likely been adapted over human evolutionary history (Bleske & Buss, that people’s expectations of others can be met leading to interper- 2001). In fact, men are more motivated to initiate friendships with sonal liking when non-verbal cues are lacking. When the availabil- a member of the other sex based on sexual attraction and romantic ity of visual cues is subsumed into the desirability of physical interest than are women (Buss, 1989; Sprecher, 1989). attractiveness for relationships between members of the same Early studies of relationship formation during FtF communica- and opposite-sex, we propose the following hypothesis: tion emphasize physical attractiveness in heterosexual mate H3c: Evaluators will be more likely to initiate friendships with choice and research demonstrates that physical attractiveness is opposite-sex profile owners with no-photos than with no-photo a strong predictor of dating desirability for both sexes (Buss, same-sex dyads. 1989; Feingold, 1990; Walster, Aronson, Abrahams, & Rottman, 1966; Gangestad & Buss, 1993).This raises another question about relationship development in the context of CMC: what type of rela- 4. Method tionship is most likely to develop online? Parks and Roberts (1998) studied several virtual communities and found that opposite-sex 4.1. Participants relationships dominated friendships in cyberspace and these on- line relationships did not differ from relationships developed off- A convenience sample of college students in an introductory line. The extent that men and women may differ in the communication course (COM101) at a public university in the east- importance they accord to physical attractiveness allows us to ap- ern United States was recruited and participants were granted re- proach this issue empirically by adding gender as a predictor of search credit for their voluntary participation. An announcement inferences about their online communicative partners. Thus, was made in class to approximately 475 students by one of our re- H2: There will be an interaction effect between profile owner’s search team members calling for their participation. A total of 350 gender and evaluator’s gender such that willingness to initiate students completed the survey, yielding roughly a 73.7% response friendship will be higher among opposite-sex profile owners across rate. all three conditions (no-photo-attractive photo-unattractive Participants (93%) were randomly assigned to one of the six photo). experimental conditions. However, because of the limited sample Drawing on the possible majority of opposite-sex relationships pool, towards the end a few participants were forced into specific formed online (Parks & Roberts, 1998), we further hypothesize conditions to ensure equal cell sizes. About 11 percent (39 partic- H2a: Male evaluators will be more willing to initiate friendships ipants) had no Facebook account or did not indicate their gender, with female profile owners than with male profile owners. and were excluded. Male participants comprised 57% of the sam- H2b: Female evaluators will be more willing to initiate friend- ple. The mean age for the sample was 20.2 years (SD = .187); par- ships with male profile owners than with female profile owners. ticipants ranged from 18 to 36 years old. The number of People search for same-sex friends similar to themselves based respondents in each group for the final sample is shown in Table 1. on a variety of dimensions such as intelligence, physical attractive- ness, social class, educational level, and age (Buss, 1984). While same-sex friendships provide adaptive benefits, they may also be 4.2. Procedure costly and destructive. For example, strong similarity between same-sex friends may result in competing for or stealing mates A link was posted on the shared course website by the instruc- (Bleske & Buss, 2000; Buss, 1984; Tolson & Urberg, 1993). tor so the participants had access to the online survey anytime Kernis and Wheeler (1981) further distinguish ‘‘radiation” and they wished. This single URL directed them to a consent form ‘‘contrast” effects: people are perceived more attractively when which explained that the study was to investigate friendship for- they are associated with physically attractive friends of the oppo- mation on Facebook and they would be asked to consider a ficti- site-sex. This radiation effect is consistent with the what is beautiful tious person’s profile. Participants were then exposed to one of is good stereotype (Dion et al., 1972). Contrast effects (Kernis & the six prepared Facebook-like profiles. An online questionnaire Wheeler, 1981) suggest that people are rated more negatively measuring their willingness to initiate a friendship with the profile when in the presence of attractive same-sex friends than unattrac- owner followed whichever profile they viewed. After rating their tive ones. Hence, highly attractive people experience the highest willingness to initiate a friendship, demographic information rejection rate during social contacts with same-sex peers (Krebs (e.g., age, gender, and college year) and other basic information & Adinolfi, 1975). about their Facebook use was collected. 230 S.S. Wang et al. / Computers in Human Behavior 26 (2010) 226–234
Table 1 4.3.2. Willingness to initiate friendship Means and standard deviations. Four Likert-type items were used to measure whether respon- Attractive Unattractive No-photo dents were willing to accept fictitious profile owners as friends Male Female Male Female Male Female and further interact with them online. These items were designed to reflect the intention to engage in online exchanges beyond the Evaluator Male initial, low cost behavior of accepting the other as a Facebook M 2.24 3.66 2.29 2.13 2.09 3.13 friend. These items included whether or not participants were will- SD 1.06 1.28 1.15 1.05 1.08 1.65 ing to add the profile owner as a friend, accept a friendship invita- N 35 29 30 30 28 26 tion from the profile owner, ‘‘poke” the profile owner, and write on Female profile owner’s wall (Cronbach’s alpha reliability score = .82). Re- M 2.92 2.04 2.08 2.54 2.23 2.16 SD 1.62 0.84 1.24 1.12 1.57 1.10 sponses ranged from 1 (strongly disagree) to 7 (strongly agree). N 20 23 15 26 27 22 An average was computed across the four items to estimate one’s ‘‘willingness to initiate friendship” score (M = 2.47, SD = 1.32). Ta- ble 1 reports standard deviations with means. To test the normality of the variable’s distribution, both skew and kurtosis were mea- sured (Skew = .861 and Kurtosis = .484) and found to be within 4.3. Measures acceptable limits. Levene’s homogeneity test was conducted to determine whether there were significant differences in variability 4.3.1. Facial attractiveness of the dependent variable between all conditions. The error of the Although past research on physical attractiveness has included dependent variable was equal across groups, indicating both facial and bodily attractiveness, facial symmetry (Jones et al., F(11, 313) = 1.72, p > .05 indicating that the variances were 2001), facial averageness, and distinctiveness (Rhodes et al., 2001; homogeneous. Langlois & Roggman, 1990; Langlois, Roggman, & Musselman, 1994; Rhodes & Tremewan, 1996) were also commonly discussed. Drawing on evolutionary psychology, facial attractiveness judg- 5. Results ments tend to be consistent across sexes, sexual orientations, eth- nic groups, and ages (Alley & Cunningham, 1991; Grammer & To examine both the main and interaction effects for visual Thornhill, 1994; Langlois & Roggman, 1990; Rhodes, Proffitt, Grady, cues, profile owner gender and evaluator gender, a 2 (owner gen- & Sumich, 1998; Thornhill & Gangestad, 1999). For example, Cunn- der: male, female) Â 3 (visual condition: attractive, unattractive, ingham, Roberts, Barbee, and Druen (1995) observed that Asian, no-photo) Â 2 (evaluator’s gender: male, female) between subjects Hispanic, Black, and White men all rated females with ‘‘high eye- analysis of variance (ANOVA) was employed. Willingness to initi- brows, widely spaced large eyes with dilated pupils, high cheek- ate friendship on Facebook was the primary dependent variable. bones, small nose, a narrow face with thin cheeks, large smile, Recall that this study is designed to explore the role visual cues full lower lip, small chin, and fuller hairstyle” as more attractive play in initial friending behavior on SNSs like Facebook. (p. 275). Facial symmetry (Grammer & Thornhill, 1994; Rhodes et al., 1998), averageness (Alley & Cunningham, 1991; Langlois & 5.1. Manipulation check Roggman, 1990), and sexual dimorphism (Thornhill & Gangestad, 1999) are also important factors in determining facial The attractiveness manipulation was tested to assure the se- attractiveness. lected photos of attractive and unattractive males and females Profile photos of attractive and unattractive males and females made specific distinctions. A total of 44 subjects were recruited were chosen based on this theory; profiles without photos were and randomly divided into four groups of equal size (N = 11). Four also included. Photos for the profiles were acquired from Braun, photos comprised of two male (attractive and unattractive) and 1 Gruendl, Marberger, and Scherber’s report (2001) and were pub- two female photos (attractive and unattractive) were randomly as- licly accessible through the website Beautycheck (http://www.bea- signed to each group, composed of raters from both sexes. Partici- utycheck.de/). This large research project on facial attractiveness pants’ perceptions of the profile owners’ facial attractiveness were was conducted at Regensburg and Rostock Universities in Ger- measured using six Likert scale items (e.g., He/she has an attractive many. The software-manipulated compound facial images are face, or he/she is sexy looking). composites from a number of real people’s photos and have been Two separate independent sample t-tests were conducted. The tested based on the facial symmetry hypotheses in a series of results suggest that all of the attractiveness manipulations were experiments. successful. The two male photos were perceived as significantly Using the facial images of Braun et al.’s report (2001) is justifi- different on the attractiveness scores, t(20) = 2.10, p < .05, r2 = .18. able as their attractiveness was previously tested based on facial The two female photos were also significantly different from each symmetry. Moreover, it was advantageous to use these images be- other in terms of the perceived attractive scores, t(20) = 4.12, cause these photos were produced by altering identical images p < .01, r2 = .46. using a computer program. Therefore, these images mitigate the possibility of biased attractiveness assessment. 5.2. Hypotheses tests Attractiveness was operationalized as a dichotomous variable and the specific photos selected were congruent with the existing H1 predicted that there would be a main effect for display of body of research. Figs. 1 and 2 shown below are sample profiles profile photos on Facebook. The results of the ANOVA indicated a with an attractive versus unattractive facial image of a female that significant main effect for visual cues on respondents’ willingness were used in our experiment.1 Other optional information available to initiate friendships, F(2, 299) = 3.51, p < .05, 2 = .02. Thus, the on typical profiles was minimized to increase internal validity. g H1 was supported. Post hoc comparisons using Tukey’s procedure at p < .05 indicated that evaluator’s viewing attractive profile own- er’s photos (M = 2.72 , SD = 1.35) had more willingness to initiate 1 Photos were acquired from Braun, C., Gruendl, M., Marberger, C. & Scherber, C. a (2001). Beautycheck – Ursachen und Folgen von Attraktivitaet. Report.
Fig. 1. Sample prepared Facebook profile with attractive female profile photo.
Fig. 2. Sample prepared Facebook profile with unattractive female profile photo.
stimulus (M = 2.40ab, SD = 1.43) did not differ from either those in female profile owners (M = 2.26, SD = 1.04), t(131) = .71, p = .71. the attractive photo condition or in the unattractive condition. Thus, results were consistent with H2a, but not with H2b. Thus, H1a was partially supported, but H2b was rejected. H3 predicted that the simple opposite-sex gender interaction H2 predicted that there would be an interaction effect between would be moderated by visual conditions. The analysis indicated profile owner’s and evaluator’s gender on willingness to initiate that there was a significant three-way interaction (profile owner friendships such that willingness to initiate friendship would be gender by evaluator gender by visual cue), F (2, 299) = 13.02, higher among opposite-sex dyads than same-sex dyads. As Fig. 3 p < .001, g2 = .05, supporting H3. The two-way interaction between shows, there was a significant interaction effect for gender, F (1, profile owner gender and evaluator gender was examined sepa- 299) = 16.20, p < .01, g2 = .03. Two separate independent sample rately for each visual cue. H3a predicted that evaluators would t-tests were conducted and indicated that male evaluators were be more likely to initiate friendships with opposite-sex profile more willing to initiate friendships with female profile owners owners with attractive photos than with same-sex dyads. As (M = 2.96, SD = 1.47) than with male profile owners (M = 2.21, shown in Fig. 4, the results showed that male evaluators exposed SD = 1.09), t(176) = 3.89, p < .01, r2 = .08, whereas for female evalu- to attractive photos were more willing to initiate friendships with ators, male profile owners (M = 2.42, SD = 1.53) did not differ from female profile owners (M = 3.66, SD = 1.28) than with male profile 232 S.S. Wang et al. / Computers in Human Behavior 26 (2010) 226–234
Fig. 3. Interaction effect between profile owner’s and evaluator’s gender. Fig. 6. Two-way interaction between profile owner gender and evaluator gender in the no-photo condition.
6. Discussion
The goal of this study was to explore the effect of visual cues on initiating friendships on Facebook. We posited physical attractive- ness of the message sender and gender of both message sender and receiver would influence the willingness to initiate friendship with zero-history CMC partners. This investigation not only addresses the theoretical significance of the hyperpersonal framework but also illustrates gender differences in the judgment of friendship
Fig. 4. Two-way interaction between profile owner gender and evaluator gender in initiation during one-shot online encounters when visual cues the attractive photo condition. are either available or absent. Display of one’s profile photo on Facebook had a significant main effect on willingness to initiate friendships with the profile owners (M = 2.24, SD = 1.05), while female evaluators exposed to owners. Physical attractiveness was most salient as a visual cue attractive photos are more willing to initiate friendships with male when choosing whom to befriend when other verbal or non-verbal profile owners (M = 2.92, SD = 1.62) than with female profile own- cues were limited. A three-way interaction of gender and appear- ers (M = 2.04, SD = .84). Two separate independent sample t-tests ance was also revealed. The results suggest that both male and fe- supported H3a. male subjects were more willing to initiate friendships with Similarly, H3c predicted that evaluators would be more willing opposite-sex profile owners with attractive photos. They also dis- to initiate friendships with opposite-sex profile owners with no- played comparatively higher willingness to make friends with photos than with same-sex dyads. As Fig. 6 shows, the results the profile owners who did not include visual cues than with those indicated that male evaluators were more willing to initiate friend- who revealed an unattractive photo. Male subjects in particular ships with female profile owners (M = 3.13, SD = 1.65) than male were found to be more willing to be friends with the visually anon- 2 profile owners (M = 2.09, SD = 1.08), t(52) = 2.78, p < .01, r = 13, ymous (no profile photo) female profile owners than with anony- when they were exposed to the no-photo conditions; female eval- mous males. uators, however, did not differ in their willingness to initiate The empirical findings of this study suggest a number of impor- friendships with male profile owners (M = 2.23, SD = 1.57) or fe- tant implications for the understanding of how people decide to male profile owners (M = 2.16, SD = 1.10) when exposed to the make friends in the virtual world. First, at a theoretical level, our no-photo condition. Contrary to H3a and H3c, H3b predicted the findings resonate with the hyperpersonal model in general; they opposite direction to the two-way interaction effect. However, indicate a gender difference when establishing dyadic relation- the two independent sample t-tests did not yield significance for ships in particular. That is, despite the persistent primacy of phys- the two-way interaction effect (see Fig. 5). Thus, data were incon- ical attractiveness on the initiation stage of relationship either sistent with the prediction. offline or online, especially with regard to opposite-sex relation- ships, visual anonymity often facilitates the relational process. This finding is consistent with the hyperpersonal model of interaction which suggests the exaggeration of impressions about CMC part- ners based on the displayed information. Exclusion of visual cues in their SNS profiles may, therefore, be one strategy of selective self-presentation for message senders. From a message receiver’s perspective, the absence of visual information is supplemented through the process of idealizing the partner based on other cues. In CMC, textual information has been regarded as the source of halo effects (Thorndike, 1920) on general attitudes toward virtual partners including their physical attractiveness. This manifestation of the affect heuristic allows CMC users to make judgments about their partners without being aware that their unconscious assump- tions play a role in their cognitive processes. As a result, the hyper- Fig. 5. Two-way interaction between profile owner gender and evaluator gender in personal effect persists in the context of SNSs where users can the unattractive photo condition. provide infinite visual cues about themselves if they so choose. S.S. Wang et al. / Computers in Human Behavior 26 (2010) 226–234 233
It nonetheless needs to be noted that the hyperpersonal model is explored with more textual information about the Facebook profile an extension of the SIP perspective (Walther, 1992). SIP predicts owners and tested in a longitudinal assessment. that verbal and paralinguistic cues lead to uncertainty reduction We also acknowledge that willingness to initiate a friendship in CMC environments. Although social cues conveyed in CMC are carries with it the possibility that the target might decline the reduced compared to FtF interaction, SIP posits that impression for- friendship request and this may inhibit people from making friend mation and relational communication can still be developed as long requests via Facebook. However, there is evidence that uninhibited as adequate time is given. The time required to construct and de- friending of strangers online is common practice. Recall on average code messages is crucial in text-based CMC. The ability to express 15% of Facebook user’s networks are comprise of people never actu- emotions in text and selective self-presentation are very important ally met FtF (Stefanone et al., 2008). Data specifically addressing for creating a hyperpersonal atmosphere over a longer period of participant’s likelihood to make contact with formerly unknown time, leading to the development of friendships. In the present others were not assessed in the current study. Yet, it is important SNS setting, however, photos individuate people quickly and effec- to note that generally participants were not likely to initiate friend- tively. The inclusion of a small but clearly identifying photo of pro- ships with relative strangers, evident by the low mean scores for file owners accompanying personal text messages stimulates this variable (M = 2.47, SD = 1.32). Moreover, given the fact that this group-level identification. This study underscores the importance study was based on potential behavior, people’s actual friending of visual cues in conveying social immediacy and forming initial decisions may yield different outcomes in real-life scenarios. impressions online. Our findings indicate that selective self-repre- sentation combined with selective reception can foster a height- References ened sense of affinity even during one-time profile exposure.
Findings of this study also suggest that the hyperpersonal effect Adams, G. R. (1982). Physical attractiveness. In A. G. Miller (Ed.), In the eye of the should be expanded to address individual-level attributes. Gender beholder: Contemporary issues in stereotyping (pp. 253–304). New York: Praeger. was a particularly significant moderator when no-photo was dis- Albada, K. F., Knapp, M. L., & Theune, K. E. (2002). Interaction appearance theory: Changing perceptions of physical attractiveness through social interaction. played in Facebook profiles. Male subjects responded especially Communication Theory, 12(1), 8–40. strongly to visual cues and were more susceptible to the hyperper- Alley, T. R. (1988). Physiognomy and social perception. In T. R. Alley (Ed.), Social and sonal effect than females. Females were more cautious about peo- applied aspects of perceiving faces (pp. 167–186). Hillsdale: Erlbaum. ple they did not know. This gender difference can be attributed to Alley, T. R., & Cunningham, M. R. (1991). Averaged faces are attractive, but very attractive faces are not average. Psychological Science, 2, 123–125. various factors such as the tendency toward impression inflation, Alley, T. R., & Hildebrandt, K. A. (1988). Determinants and consequences of facial the perceived importance of physical attractiveness in virtual aesthetics. In T. R. Alley (Ed.), Social and applied aspects of perceiving faces friendship formation, and potentially different relational seeking (pp. 101–140). Hillsdale: Erlbaum. Ambady, N., & Rosenthal, R. (1992). Thin slices of behavior as predictors of purposes in Facebook use. interpersonal consequences: A meta-analysis. Psychological Bulletin, 2, 256–274. A second implication of our findings is that recent SNS settings Argyle, M., & Dean, J. (1965). Eye-contact, distance and affiliation. Sociometry, 28, like Facebook can foster social norms for social stereotyping con- 289–304. Bell, R. (1981). Friendships of women and men. Psychology of Women Quarterly, 5, struction. Earlier research on CMC found that stereotypical re- 402–417. sponses were possible even when cues were limited to a Berscheid, E., & Walster, E. H. (1974). Physical attractiveness. In L. Berkowitz (Ed.). feminine or masculine first name (Matheson & Zanna, 1990). Our Advances in experimental social psychology (Vol. 7, pp. 158–215). New York: Academic Press. findings support this proposition as physical attractiveness and Bleske, A. L., & Buss, D. M. (2000). Can men and women be just friends? Personal gender stereotypes were associated with a pattern by which more Relationships, 7, 131–151. attractive people were viewed more favorably as virtual friends Bleske, A. L., & Buss, D. M. (2001). Opposite-sex friendship: Sex differences and similarities in initiation, selection, and dissolution. Personality and Social and opposite-sex friendships were more likely to be formed even Psychology Bulletin, 27, 1310–1327. in the no-photo condition. The view that this form of CMC lacks Buss, D. M. (1984). Evolutionary biology and personality psychology: Toward a adequate social cues was not supported; indeed socially recogniz- conception of human nature and individual differences. American Psychologist, able meanings, even given cues selectively presented in the time- 39, 361–377. Buss, D. M. (1989). Sex differences in human mate preferences: Evolutionary limited situation, were observed. hypotheses tested in 37 cultures. Behavioral and Brain Sciences, 12, 1–49. Finally, visual cues can play a critical role during social interac- comScore (2009). Top 50 US Web Properties for August 2009. tion online and offline alike. Physical attractiveness is one of the
Feingold, A. (1992). Good looking people are not what we think. Psychological Patzer, G. L. (1985). The physical attractiveness phenomena. New York: Plenum Press. Bulletin, 11, 304–341. Peluchette, J., & Karl, K. (2008). Social networking profiles: An examination of Gangestad, S. W., & Buss, D. M. (1993). Pathogen prevalence and human mate student attitudes regarding use and appropriateness of content. preferences. Ethology and Sociobiology, 14, 89–96. Cyberpsychology & Behavior, 11(1), 95–97. Goffman, E. (1963). Behavior in public places: Notes on the social organization of Reingold, H. (1993). The virtual community. New York: Addison-Wesley. gathering. New York: Free Press. Rhodes, G., Proffitt, F., Grady, J. M., & Sumich, A. (1998). Facial symmetry and the Grammer, K., & Thornhill, R. (1994). Human (Homo sapiens) facial attractiveness and perception of beauty. Psychonomic Bulletin & Review, 5(4), 659–669. sexual selection: The role of symmetry and averageness. Journal of Comparative Rhodes, G., & Tremewan, T. (1996). Averageness, exaggeration and facial Psychology, 108, 233–242. attractiveness. Psychological Science, 7, 105–110. Gross, R., & Acquisti, A. (2005). Information revelation and privacy in online social Rhodes, G., Zebrowitz, L., Clark, A., Kalick, S., Hightower, A., & McKay, R. (2001). Do networks. In Proceedings of ACM’s twelfth annual workshop on privacy in the facial averageness and symmetry signal health? Evolution and Human Behavior, electronic society (pp.71–80). New York, NY. 22, 31–46. Hacker, H. M. (1981). Blabbermouths and clams: Sex differences in self-disclosure Riggio, R. E. (1986). Assessment of basic social skills. Journal of Personality and Social in same-sex and cross-sex friendship dyads. Psychology of Women Quarterly, 5, Psychology, 51, 649–660. 385–401. Roszell, P., Kennedy, D., & Grabb, E. (1989). Physical attractiveness and income Hancock, J., & Dunham, P. (2001). Impression formation in computer-mediated attainment among Canadians. Journal of Psychology, 123, 547–559. communication revisited: An analysis of the breadth and intensity of Sapadin, L. A. (1988). Friendship and gender: Perspective of professional men and impression. Communication Research, 28, 325–332. women. Journal of Social and Personal Relationships, 5, 387–403. Hancock, J., & Toma, C. (2009). Putting your best face forward: The accuracy of Schonfeld, E. (2009). Facebook is now the forth largest site in the world. online dating photographs. Journal of Communication, 59(2), 367–386. TechCrunch.com.